The Long View | Sramana Mitra on Web 3.0 and the Science of Personalised Shopping

Sramana Mitra | Source: Sramana Mitra

MENLO PARK, United States — Since 2003, when Web 2.0 first hit the mainstream, we’ve seen an explosion of user-generated content (UGC). Since then, fashion innovators have harnessed the growth of UGC to build experiences like Polyvore, where consumers can mix and match their favourite fashion items for others to see and shop.

But as more and more consumers become increasingly active on the social web, the volume of content being created is starting to overwhelm our ability to organise and make sense of it. Amongst the thousands of sets on Polyvore, for example, how can a fashion consumer easily find the sets that are most relevant to her personal style?

But if the explosion of content associated with Web 2.0 is the primary cause of this problem, it also contains the seeds of a solution that will usher in the next phase of the internet and create an exciting opportunity for fashion retailers. In a world where people constantly share personal information, it’s becoming increasingly possible for retailers to analyse this information to better understand the specific context of the individual — her interests, personal style and other parameters — and deliver content and products that are personalised to her needs and desires. Simply put, “Web 3.0” will enable personalised experiences built on the data created by Web 2.0.

BoF spoke with Sramana Mitra, a Silicon Valley strategy consultant, author and entrepreneur — among other projects, Ms. Mitra ran Uuma, a VC-backed personalised fashion startup which received an acquisition offer from Ralph Lauren before the company was caught in the first dotcom crash — to find out more.

BoF: What is Web 3.0 and what does it mean for fashion brands?

SM: I define Web 3.0 as a verticalised, personalised user experience. The web is still utterly fragmented. You have to go to different places to find information about the same context. I have long had the vision of a personalised Saks Fifth Avenue. I want my store — my personal store — that carries merchandise that applies to me; that suits my hair colour, eye colour, skin tone, body shape and personal style. I want it to stock my favourite designers and more like those. And I want to see articles and community discussions that are specific to my interests.

I start from the notion of context and propose that all the contextual elements are made available within the same site or web service. That includes content, community, commerce, vertical search and personalisation.

BoF: Do you think Google’s effectively leverages Web 3.0? Is anyone in fashion doing this well?

SM: No, is not set up to do what I am talking about. You need an expert system powering the backend engine, not just “I like this” and “I don’t like this” type of input from consumers. Frankly, I haven’t seen any site that is really doing a good Web 3.0 implementation in fashion.

BoF: Offline, the best personal shoppers have up to the minute knowledge of fashion and can offer intelligent recommendations based on complex inputs like your personal taste or a specific event you might be shopping for. How exactly might a Web 3.0 service tackle a problem like this?

SM: The entire expert system needs to have a data structure against which the merchandise gets categorized and a rules engine that does the matchmaking based on that characterization. In other words, all the parameters — user, mood, size, event, location, time — need to be clustered, categorised and tied to a rules engine that is the personal shopper.

BoF: But in a trend-based industry like fashion, the rules are always changing!

SM: Every season, the merchandise will need to be entered into the inventory database according to the categorisation. Am I talking too much computer science? This is a complex system to build and to my knowledge, it hasn’t been built yet. At Uuma, this is what we had started to build. This was the central vision driving Uuma’s personalised store and personal shopper service.

BoF: With the coming of personal shopping agents, will Web 3.0 fundamentally democratise the delivery of personalised services that were once considered a luxury? What opportunities are there for luxury brands to differentiate their 3.0 offerings?

SM: There are degrees of personalised experiences. A basic level of personalisation can be achieved by mass brands using technology, especially intelligent agents. But luxury brands could layer in additional levels of personalization, including human personal shoppers who use technology to track your closet and your preferences and then offer additional judgment to augment the user experience and put together custom collections. My point is, the entire user experience can be massively enhanced to reach degrees of customization and personalization that we haven’t seen yet.

Finally, if you mine the data collecting in these online stores and pass some of that back to the designers, they can create custom collections based on specific types and styles of customers and be able to sell them effectively, knowing exactly what retail interfaces the customers are hanging out at. These are incredibly exciting possibilities for the business of fashion to become more personalised on the one hand — and also more scientific on the other. The fashion industry could become more financially successful by utilising personal data: analysing it and designing and merchandising accordingly.

Vikram Alexei Kansara is Managing Editor of The Business of Fashion

Related Articles

Post a Comment


  1. Awesome article! That’s why is uniquely positioned to assist consumers and marketers alike. Users can inventory their closets — even their individual outfits — to receive relevant products recommendations, which may be purchased and/or shared with their social network.

  2. Welcome Web 3.0! We Need It!

    Web 2.0 Wrong’s-Filters: Not all have SIZE, options many times out-of-date no longer in stock, despite many filters…still hard to find what your looking for, labels-hard to distinguish brands different lines ex. Looking for Micheal Kors only-also get Michael by Michael Kors, MICHAEL Michael Kors, Michael Michael Kors

  3. ‘her’, ‘her’, ‘her’… I guess ‘he’ no longer indulges in online shopping or social media. What Sramana is proposing is also counter-productive. Having everything fine-tuned on a commerce site will duplicate clones of one another, it limits the range or provocation to purchase outside sequential shopping habits (which is not clever at all). Also relaying feedback to design studios compromises the designer and undermines the entire process. Ye it might improve net profit margins, brilliant for the shareholders, what about the consumer experience? We don’t want to shop the same model of everything, everywhere, just because a piece of software says so. This model of web 3.0 removes any form of empathic involvement from brands and does not tolerate intuitive decision making. What’s next? No need for buyers?

    moi from Yeovil, Somerset, United Kingdom
  4. Forgot to mention, web 3.0 should be proposing an articulated question to the consumer, not bombarding them with sameness. This is why in my opinion google adverts are 100% useless. They simply take note of where you have been browsing, then throw up purchasing choices of products that you have already (previously) dismissed while on that site. It’s one-step behind the consumer…

    moi from Yeovil, Somerset, United Kingdom
  5. Although I thought and visioned similar technology that can offer much deeper personalized services, I also come to think that might restrict one’s experience. I believe experiencing not-expected nor planned always add great insight and fun to our lives. Just a thought to add.

    july from United States
  6. This is just what the fashion world needs. I have had exactly the same vision but it is a complex proces and it needs to be excuted very precise. I have seen boutiques, myshape and countless others but the fact of the matter is they are not implementing this vision right.

    Anyone who is working on this subject and mabye is putting a team together to accomplisch this vision feel free to send me an email at

    An from Den Haag, South Holland, Netherlands
  7. The explanation above of Web 3.0 sounds different to what the industry refers to as Web 3.0 is the Semantic Web, structured content, linked data and microformats that are machine readable and will help aggregators read the data from sites without the need for complicated web-scraping tools. Intelligent services can be created to consume Web 3.0 data.

    One of the biggest issues we’re face with is categorisation of this product data, taxonomies have many problems context,ambiguity, fragmentation etc. Folksonomies have even bigger problems because of the UGC element, not to mention localisation of these.

    Currently price comparison engines like Google Shopping rely on specifically generated XML from retailers to list prices and descriptions, it’s not a closed standard but not easily accessible to public services.

    Linking or ‘clustering’ products etc in a taxonomy is not new, it’s based on a mathematical principle called singular vector decompositioning.
    It’s not unlike measuring the distance between objects, their distance to surrnoding objects determines how ‘related’ they are. In a practical implementation this is termed Latent Sematic Indexing.

    As for some peoples argument that aggregators like this will create a homogenised view, this could be true for a weak recommendation service, but not on the scale being suggested. Entropy and serendipity is an important part of exploration and recomendation process. New forms of passive measuring interest will be developed, for instance utilising your webcam to eyetrack which product on your screen draw your attention, how you navigate etc. Observation is key, behavioural economics has taught us that questions (or ratings) affect our implicit reactions and need to be avoided.

    These personalisation engines will be used for everything, content, products, food, music. Generally tech in fashion industry is a little slow on the uptake so we’ll see better implementations in other industries first.

    To see what one of the more impressive recommendation engines is doing right now check out, Chris Dixons, Hunch.

  8. While I see how feeding data back to designers could aid in the product development process for the designer; I tend to think that it makes that process less organic. Fashion is about creating art and shaping perspectives. And yes, it is an industry that thrives on turning a profit as well. But I truly believe that it loses some of its magic when the creative vision is infiltrated by the thoughts and opinions of the people who will consume it. The designer begins to take on the role of order taker, rather than visionary.

  9. “I want my store — my personal store — that carries merchandise that applies to me; that suits my hair colour, eye colour, skin tone, body shape and personal style. ”

    Who is she referring to? I certainly wouldn’t want that. The whole concept of having a website tell you exactly what you should buy is actually quite sad, it seems that this would not in fact serve the customer, it would serve the brands whose product was “presented” to the customers, while the customers themselves become totally homogenized. And what would happen to smaller brands and independent boutiques? What about the thrill of shopping, the excitement of a discovery that wasn’t presented to you by some website?

    I’d like to know what the backlash to this will be, as there is bound to be one. In fact, this might fuel a very interesting rebellion in the fashion industry.

  10. I don’t see why everyone is complaining about this. The average clothing consumer has a general style that they will not stray from, whether or not there is a personalized “store”. This system would actually, most likely, increase the the number of products a consumer will be exposed to that they actually like. It would just decrease the dross you have to wade through to find what you like. Those with more eccentric and unique styles do not HAVE to use the personalized sites, it’s just an alternative.

    Kiah Leigh from Pasadena, MD, United States
  11. Right. I am also missing the logic of why a personalized store is in conflict with personalized style. More personalization data in the hands of designers would mean a higher degree of precision in custom designing for different types / styles of consumers … and yes, a personalized store would make it easier to find things that you like.

    Question about the definition of Web 3.0 above would perhaps get resolved by referring to this post: Web 3.0 and the Semantic Web :

    Finally, those who share the vision, thank you for your enthusiasm. If some of you are fashion e-commerce entrepreneurs trying to build businesses in this domain, please consider joining the One Million by One Million program aimed at helping a million entrepreneurs reach a million dollars in annual revenue and beyond. [] We would love to have more fashion entrepreneurs in the mix.

  12. I’m not sure why there’s a fear of homogenization out of personalization.

    >>>”Who is she referring to? I certainly wouldn’t want that. The whole concept of having a website tell you exactly what you should buy is actually quite sad.” <<<
    I'd imagine quite a few people would like to be shown things they could be interested in ahead of things that weren't suitable. If you never wear bodycon dresses and favor the style of Rick Owens and Alexander Wang, you probably wouldn't be interested in seeing the latest Herve Leger or Versace dress. And if someone is the Rick Owens type, hammering them with items from Versace doesn't help stores or customers.

  13. For the past year the team at Dressipi have been working on a similar
    service: a personalised vertical search service, that acts as a
    personal shopper/stylist for our users.

    Personalising something as complex as fashion has required blending
    the talents of a team of engineers, statisticians and fashion
    stylists. As Alastair points out there are many issues to solve in
    terms of data, context, emotion, ontology etc. Our approach has been to create our own taxonomy, and capture the expertise of fashion stylists as executable rules. This has taken us a year to get right, and we’re just scratching the surface of the problem, as we begin to better handle the subtle dependencies and emotional considerations of shopping and dressing.

    We’ve been in beta for the last 8 months with around 1500 people, and about to open up to the public in the next 2 months. We’d be happy to chat to anyone interested in understanding more. We’re based in London, UK.